Merge branch 'dash' into dash_acl

Author: Pterosaur

cfgmgr/Makefile.am

@@ -15,7 +15,10 @@ dist_cfgmgr_DATA = \
 		buffer_pool_mellanox.lua \
 		buffer_check_headroom_vs.lua \
 		buffer_headroom_vs.lua \
-		buffer_pool_vs.lua
+		buffer_pool_vs.lua \
+		buffer_check_headroom_barefoot.lua \
+		buffer_headroom_barefoot.lua \
+		buffer_pool_barefoot.lua
 
 if DEBUG
 DBGFLAGS = -ggdb -DDEBUG

cfgmgr/buffer_check_headroom_barefoot.lua

@@ -0,0 +1,6 @@
+local ret = {}
+
+table.insert(ret, "result:true")
+table.insert(ret, "debug:No need to check port headroom limit as shared headroom pool model is supported.")
+
+return ret

cfgmgr/buffer_headroom_barefoot.lua

@@ -0,0 +1,147 @@
+-- KEYS - profile name
+-- ARGV[1] - port speed
+-- ARGV[2] - cable length
+-- ARGV[3] - port mtu
+-- ARGV[4] - gearbox delay
+
+-- Parameters retried from databases:
+-- From CONFIG_DB.LOSSLESS_TRAFFIC_PATTERN
+--      small packet percentage:    the parameter which is used to control worst case regarding the cell utilization
+--      mtu:                        the mtu of lossless packet
+-- From STATE_DB.ASIC_TABLE:
+--      cell size:                  cell_size of the ASIC
+--      pipeline_latency:           the latency (XON)
+--      mac_phy_delay:              the bytes held in the switch chip's egress pipeline and PHY when XOFF has been generated
+--      peer_response_time:         the bytes that are held in the peer switch's pipeline and will be send out when the XOFF packet is received
+
+local lossless_mtu
+local small_packet_percentage
+local cell_size
+local pipeline_latency
+local mac_phy_delay
+local peer_response_time
+
+local port_speed = tonumber(ARGV[1])
+local cable_length = tonumber(string.sub(ARGV[2], 1, -2))
+local port_mtu = tonumber(ARGV[3])
+local gearbox_delay = tonumber(ARGV[4])
+
+local config_db = "4"
+local state_db = "6"
+
+local ret = {}
+
+-- Pause quanta should be taken for each operating speed is defined in IEEE 802.3 31B.3.7.
+-- The key of table pause_quanta_per_speed is operating speed at Mb/s.
+-- The value of table pause_quanta_per_speed is the number of pause_quanta.
+local pause_quanta_per_speed = {}
+pause_quanta_per_speed[400000] = 905
+pause_quanta_per_speed[200000] = 453
+pause_quanta_per_speed[100000] = 394
+pause_quanta_per_speed[50000] = 147
+pause_quanta_per_speed[40000] = 118
+pause_quanta_per_speed[25000] = 80
+pause_quanta_per_speed[10000] = 67
+pause_quanta_per_speed[1000] = 2
+pause_quanta_per_speed[100] = 1
+
+-- Get pause_quanta from the pause_quanta_per_speed table
+local pause_quanta = pause_quanta_per_speed[port_speed]
+
+if gearbox_delay == nil then
+    gearbox_delay = 0
+end
+
+-- Fetch ASIC info from ASIC table in STATE_DB
+redis.call("SELECT", state_db)
+local asic_keys = redis.call("KEYS", "ASIC_TABLE*")
+
+-- Only one key should exist
+local asic_table_content = redis.call("HGETALL", asic_keys[1])
+
+for i = 1, #asic_table_content, 2 do
+    if asic_table_content[i] == "cell_size" then
+        cell_size = tonumber(asic_table_content[i+1])
+    end
+    if asic_table_content[i] == "pipeline_latency" then
+        pipeline_latency = tonumber(asic_table_content[i+1]) * 1024
+    end
+    if asic_table_content[i] == "mac_phy_delay" then
+        mac_phy_delay = tonumber(asic_table_content[i+1]) * 1024
+    end
+    -- If failed to get pause_quanta from the table, then use the default peer_response_time stored in state_db
+    if asic_table_content[i] == "peer_response_time" and  pause_quanta == nil then
+        peer_response_time = tonumber(asic_table_content[i+1]) * 1024
+    end
+end
+
+-- Fetch lossless traffic info from CONFIG_DB
+redis.call("SELECT", config_db)
+local lossless_traffic_keys = redis.call("KEYS", "LOSSLESS_TRAFFIC_PATTERN*")
+
+-- Only one key should exist
+local lossless_traffic_table_content = redis.call("HGETALL", lossless_traffic_keys[1])
+for i = 1, #lossless_traffic_table_content, 2 do
+    if lossless_traffic_table_content[i] == "mtu" then
+        lossless_mtu = tonumber(lossless_traffic_table_content[i+1])
+    end
+    if lossless_traffic_table_content[i] == "small_packet_percentage" then
+        small_packet_percentage = tonumber(lossless_traffic_table_content[i+1])
+    end
+end
+
+-- Fetch the shared headroom pool size
+local shp_size = tonumber(redis.call("HGET", "BUFFER_POOL|ingress_lossless_pool", "xoff"))
+
+-- Calculate the headroom information
+local speed_of_light = 198000000
+local minimal_packet_size = 64
+local cell_occupancy
+local worst_case_factor
+local propagation_delay
+local bytes_on_cable
+local bytes_on_gearbox
+local xoff_value
+local xon_value
+local headroom_size
+
+if cell_size > 2 * minimal_packet_size then
+    worst_case_factor = cell_size / minimal_packet_size
+else
+    worst_case_factor = (2 * cell_size) / (1 + cell_size)
+end
+
+cell_occupancy = (100 - small_packet_percentage + small_packet_percentage * worst_case_factor) / 100
+
+if (gearbox_delay == 0) then
+    bytes_on_gearbox = 0
+else
+    bytes_on_gearbox = port_speed * gearbox_delay / (8 * 1024)
+end
+
+-- If successfully get pause_quanta from the table, then calculate peer_response_time from it
+if pause_quanta ~= nil then
+    peer_response_time = (pause_quanta) * 512 / 8
+end
+
+if port_speed == 400000 then
+    peer_response_time = 2 * peer_response_time
+end
+
+bytes_on_cable = 2 * cable_length * port_speed * 1000000000 / speed_of_light / (8 * 1024)
+propagation_delay = port_mtu + bytes_on_cable + 2 * bytes_on_gearbox + mac_phy_delay + peer_response_time
+
+-- Calculate the xoff and xon and then round up at 1024 bytes
+xoff_value = lossless_mtu + propagation_delay * cell_occupancy
+xoff_value = math.ceil(xoff_value / 1024) * 1024
+xon_value = pipeline_latency
+xon_value = math.ceil(xon_value / 1024) * 1024
+
+headroom_size = xon_value
+headroom_size = math.ceil(headroom_size / 1024) * 1024
+
+table.insert(ret, "xon" .. ":" .. math.ceil(xon_value))
+table.insert(ret, "xoff" .. ":" .. math.ceil(xoff_value))
+table.insert(ret, "size" .. ":" .. math.ceil(headroom_size))
+
+return ret

cfgmgr/buffer_pool_barefoot.lua

@@ -0,0 +1,30 @@
+-- KEYS - None
+-- ARGV - None
+
+local result = {}
+local config_db = "4"
+local state_db = "6"
+
+redis.call("SELECT", state_db)
+local asic_keys = redis.call("KEYS", "ASIC_TABLE*")
+local cell_size = tonumber(redis.call("HGET", asic_keys[1], "cell_size"))
+
+-- Based on cell_size, calculate singular headroom
+local ppg_headroom = 400 * cell_size
+
+redis.call("SELECT", config_db)
+local ports = redis.call("KEYS", "PORT|*")
+local ports_num = #ports
+
+-- 2 PPGs per port, 70% of possible maximum value.
+local shp_size = math.ceil(ports_num * 2 * ppg_headroom * 0.7)
+
+local ingress_lossless_pool_size_fixed = tonumber(redis.call('HGET', 'BUFFER_POOL|ingress_lossless_pool', 'size'))
+local ingress_lossy_pool_size_fixed = tonumber(redis.call('HGET', 'BUFFER_POOL|ingress_lossy_pool', 'size'))
+local egress_lossy_pool_size_fixed = tonumber(redis.call('HGET', 'BUFFER_POOL|egress_lossy_pool', 'size'))
+
+table.insert(result, "ingress_lossless_pool" .. ":" .. ingress_lossless_pool_size_fixed .. ":" .. shp_size)
+table.insert(result, "ingress_lossy_pool" .. ":" .. ingress_lossy_pool_size_fixed)
+table.insert(result, "egress_lossy_pool" .. ":" .. egress_lossy_pool_size_fixed)
+
+return result

cfgmgr/buffermgr.cpp

@@ -196,7 +196,7 @@ task_process_status BufferMgr::doSpeedUpdateTask(string port)
     // Although we have up to 8 PGs for now, the range to check is expanded to 32 support more PGs
     set<string> lossless_pg_combinations = generateIdListFromMap(lossless_pg_id, sizeof(lossless_pg_id));
 
-    if (m_portStatusLookup[port] == "down" && m_platform == "mellanox")
+    if (m_portStatusLookup[port] == "down" && (m_platform == "mellanox" || m_platform == "barefoot"))
     {
         for (auto lossless_pg : lossless_pg_combinations)
         {

cfgmgr/vlanmgr.cpp

@@ -134,6 +134,11 @@ bool VlanMgr::addHostVlan(int vlan_id)
     std::string res;
     EXEC_WITH_ERROR_THROW(cmds, res);
 
+    res.clear();
+    const std::string echo_cmd = std::string("")
+      + ECHO_CMD + " 0 > /proc/sys/net/ipv4/conf/" + VLAN_PREFIX + std::to_string(vlan_id) + "/arp_evict_nocarrier";
+    swss::exec(echo_cmd, res);
+
     return true;
 }
 

orchagent/fabricportsorch.cpp

@@ -11,17 +11,18 @@
 #include "timer.h"
 
 #define FABRIC_POLLING_INTERVAL_DEFAULT   (30)
+#define FABRIC_PORT_PREFIX    "PORT"
 #define FABRIC_PORT_ERROR     0
 #define FABRIC_PORT_SUCCESS   1
 #define FABRIC_PORT_STAT_COUNTER_FLEX_COUNTER_GROUP         "FABRIC_PORT_STAT_COUNTER"
 #define FABRIC_PORT_STAT_FLEX_COUNTER_POLLING_INTERVAL_MS   10000
 #define FABRIC_QUEUE_STAT_COUNTER_FLEX_COUNTER_GROUP        "FABRIC_QUEUE_STAT_COUNTER"
 #define FABRIC_QUEUE_STAT_FLEX_COUNTER_POLLING_INTERVAL_MS  100000
-#define FABRIC_PORT_TABLE "FABRIC_PORT_TABLE"
 
 extern sai_object_id_t gSwitchId;
 extern sai_switch_api_t *sai_switch_api;
 extern sai_port_api_t *sai_port_api;
+extern sai_queue_api_t *sai_queue_api;
 
 const vector<sai_port_stat_t> port_stat_ids =
 {
@@ -42,7 +43,8 @@ static const vector<sai_queue_stat_t> queue_stat_ids =
     SAI_QUEUE_STAT_CURR_OCCUPANCY_LEVEL,
 };
 
-FabricPortsOrch::FabricPortsOrch(DBConnector *appl_db, vector<table_name_with_pri_t> &tableNames) :
+FabricPortsOrch::FabricPortsOrch(DBConnector *appl_db, vector<table_name_with_pri_t> &tableNames,
+                                 bool fabricPortStatEnabled, bool fabricQueueStatEnabled) :
         Orch(appl_db, tableNames),
         port_stat_manager(FABRIC_PORT_STAT_COUNTER_FLEX_COUNTER_GROUP, StatsMode::READ,
                           FABRIC_PORT_STAT_FLEX_COUNTER_POLLING_INTERVAL_MS, true),
@@ -55,14 +57,17 @@ FabricPortsOrch::FabricPortsOrch(DBConnector *appl_db, vector<table_name_with_pr
     SWSS_LOG_NOTICE( "FabricPortsOrch constructor" );
 
     m_state_db = shared_ptr<DBConnector>(new DBConnector("STATE_DB", 0));
-    m_stateTable = unique_ptr<Table>(new Table(m_state_db.get(), FABRIC_PORT_TABLE));
+    m_stateTable = unique_ptr<Table>(new Table(m_state_db.get(), APP_FABRIC_PORT_TABLE_NAME));
 
     m_counter_db = shared_ptr<DBConnector>(new DBConnector("COUNTERS_DB", 0));
-    m_laneQueueCounterTable = unique_ptr<Table>(new Table(m_counter_db.get(), COUNTERS_QUEUE_NAME_MAP));
-    m_lanePortCounterTable = unique_ptr<Table>(new Table(m_counter_db.get(), COUNTERS_QUEUE_PORT_MAP));
+    m_portNameQueueCounterTable = unique_ptr<Table>(new Table(m_counter_db.get(), COUNTERS_FABRIC_QUEUE_NAME_MAP));
+    m_portNamePortCounterTable = unique_ptr<Table>(new Table(m_counter_db.get(), COUNTERS_FABRIC_PORT_NAME_MAP));
 
     m_flex_db = shared_ptr<DBConnector>(new DBConnector("FLEX_COUNTER_DB", 0));
-    m_flexCounterTable = unique_ptr<ProducerTable>(new ProducerTable(m_flex_db.get(), FABRIC_PORT_TABLE));
+    m_flexCounterTable = unique_ptr<ProducerTable>(new ProducerTable(m_flex_db.get(), APP_FABRIC_PORT_TABLE_NAME));
+
+    m_fabricPortStatEnabled = fabricPortStatEnabled;
+    m_fabricQueueStatEnabled = fabricQueueStatEnabled;
 
     getFabricPortList();
 
@@ -147,32 +152,96 @@ bool FabricPortsOrch::allPortsReady()
 
 void FabricPortsOrch::generatePortStats()
 {
-    // FIX_ME: This function installs flex counters for port stats
-    // on fabric ports for fabric asics and voq asics (that connect
-    // to fabric asics via fabric ports). These counters will be
-    // installed in FLEX_COUNTER_DB, and queried by syncd and updated
-    // to COUNTERS_DB.
-    // However, currently BCM SAI doesn't update its code to query
-    // port stats (metrics in list port_stat_ids) yet.
-    // Also, BCM sets too low value for "Max logical port count" (256),
-    // causing syncd to crash on voq asics that now include regular front
-    // panel ports, fabric ports, and multiple logical ports.
-    // So, this function will just do nothing for now, and we will readd
-    // code to install port stats counters when BCM completely supports.
+    if (!m_fabricPortStatEnabled) return;
+
+    SWSS_LOG_NOTICE("Generate fabric port stats");
+
+    vector<FieldValueTuple> portNamePortCounterMap;
+    for (auto p : m_fabricLanePortMap)
+    {
+        int lane = p.first;
+        sai_object_id_t port = p.second;
+
+        std::ostringstream portName;
+        portName << FABRIC_PORT_PREFIX << lane;
+        portNamePortCounterMap.emplace_back(portName.str(), sai_serialize_object_id(port));
+
+        // Install flex counters for port stats
+        std::unordered_set<std::string> counter_stats;
+        for (const auto& it: port_stat_ids)
+        {
+            counter_stats.emplace(sai_serialize_port_stat(it));
+        }
+        port_stat_manager.setCounterIdList(port, CounterType::PORT, counter_stats);
+    }
+    m_portNamePortCounterTable->set("", portNamePortCounterMap);
 }
 
 void FabricPortsOrch::generateQueueStats()
 {
+    if (!m_fabricQueueStatEnabled) return;
     if (m_isQueueStatsGenerated) return;
     if (!m_getFabricPortListDone) return;
 
-    // FIX_ME: Similar to generatePortStats(), generateQueueStats() installs
-    // flex counters for queue stats on fabric ports for fabric asics and voq asics.
-    // However, currently BCM SAI doesn't fully support queue stats query.
-    // Query on queue type and index is not supported for fabric asics while
-    // voq asics are not completely supported.
-    // So, this function will just do nothing for now, and we will readd
-    // code to install queue stats counters when BCM completely supports.
+    SWSS_LOG_NOTICE("Generate queue map for fabric ports");
+
+    sai_status_t status;
+    sai_attribute_t attr;
+
+    for (auto p : m_fabricLanePortMap)
+    {
+        int lane = p.first;
+        sai_object_id_t port = p.second;
+
+        // Each serdes has some pipes (queues) for unicast and multicast.
+        // But normally fabric serdes uses only one pipe.
+        attr.id = SAI_PORT_ATTR_QOS_NUMBER_OF_QUEUES;
+        status = sai_port_api->get_port_attribute(port, 1, &attr);
+        if (status != SAI_STATUS_SUCCESS)
+        {
+            throw runtime_error("FabricPortsOrch get port queue number failure");
+        }
+        int num_queues = attr.value.u32;
+
+        if (num_queues > 0)
+        {
+            vector<sai_object_id_t> m_queue_ids;
+            m_queue_ids.resize(num_queues);
+
+            attr.id = SAI_PORT_ATTR_QOS_QUEUE_LIST;
+            attr.value.objlist.count = (uint32_t) num_queues;
+            attr.value.objlist.list = m_queue_ids.data();
+
+            status = sai_port_api->get_port_attribute(port, 1, &attr);
+            if (status != SAI_STATUS_SUCCESS)
+            {
+                throw runtime_error("FabricPortsOrch get port queue list failure");
+            }
+
+            // Maintain queue map and install flex counters for queue stats
+            vector<FieldValueTuple> portNameQueueMap;
+
+            // Fabric serdes queue type is SAI_QUEUE_TYPE_FABRIC_TX. Since we always
+            // maintain only one queue for fabric serdes, m_queue_ids size is 1.
+            // And so, there is no need to query  SAI_QUEUE_ATTR_TYPE and SAI_QUEUE_ATTR_INDEX
+            // for queue. Actually, SAI does not support query these attributes on fabric serdes.
+            int queueIndex = 0;
+            std::ostringstream portName;
+            portName << FABRIC_PORT_PREFIX << lane << ":" << queueIndex;
+            const auto queue = sai_serialize_object_id(m_queue_ids[queueIndex]);
+            portNameQueueMap.emplace_back(portName.str(), queue);
+
+            // We collect queue counters like occupancy level
+            std::unordered_set<string> counter_stats;
+            for (const auto& it: queue_stat_ids)
+            {
+                counter_stats.emplace(sai_serialize_queue_stat(it));
+            }
+            queue_stat_manager.setCounterIdList(m_queue_ids[queueIndex], CounterType::QUEUE, counter_stats);
+
+            m_portNameQueueCounterTable->set("", portNameQueueMap);
+        }
+    }
 
     m_isQueueStatsGenerated = true;
 }
@@ -199,7 +268,7 @@ void FabricPortsOrch::updateFabricPortState()
         int lane = p.first;
         sai_object_id_t port = p.second;
 
-        string key = "PORT" + to_string(lane);
+        string key = FABRIC_PORT_PREFIX + to_string(lane);
         std::vector<FieldValueTuple> values;
         uint32_t remote_peer;
         uint32_t remote_port;